MicrosoftAudio modelPreview
MAI-Transcribe-1.5
PopularityRank 670 of 875
0out of 100same as last month
Summary
MAI-Transcribe-1.5 is Microsoft’s speech-to-text model for 43 languages. Keyword biasing helps with domain terms, and Microsoft cites up to five times the speed of comparable models.
The endpoint accepts audio and returns text. Foundry access is a public preview without an SLA. Microsoft lists $0.36 per hour of audio. It is intended for transcription, not speech synthesis or audio-to-audio dialogue.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 45 published benchmark results
Page 1 of 4
Measurements without matching peer values
These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.
Artificial Analysis AA-WER v2.0 (non-streaming) (Show measurement, test conditions, and source)
- Source value
- 2.4
- Score
- 2.4
- Metric
- duration-weighted word error rate across AA-AgentTalk, VoxPopuli-Cleaned-AA, and Earnings22-Cleaned-AA
- Unit
- %
- Benchmark version
- AA-WER v2.0, current non-streaming board retrieved 2026-10-16
- Category
- audio
- Direction
- Lower is better
- Source type
- independent-evaluator
- Evaluator
- Artificial Analysis
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. Current board value for exact MAI-Transcribe-1.5 row. AA-WER v2 weights roughly eight hours of audio by duration: AA-AgentTalk 50%, VoxPopuli-Cleaned-AA 25%, Earnings22-Cleaned-AA 25%.
- Context
- This current direct evaluator observation is retained separately from the older vendor-announcement observation. It is not a FLEURS result; the exact historical correction is handled in the prior record.
Artificial Analysis AA-AgentTalk WER (Show measurement, test conditions, and source)
- Source value
- 2
- Score
- 2
- Metric
- word error rate on AA-AgentTalk
- Unit
- %
- Benchmark version
- AA-WER v2.0 per-dataset comparison, retrieved 2026-10-16
- Category
- audio
- Direction
- Lower is better
- Source type
- independent-evaluator
- Evaluator
- Artificial Analysis
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. Exact MAI-Transcribe-1.5 row on the AA-AgentTalk dataset; this is one component, not the overall AA-WER result.
- Context
- The article does not report a per-row sample count. This dataset-specific result is not the overall weighted AA-WER score.
Artificial Analysis VoxPopuli-Cleaned-AA WER (Show measurement, test conditions, and source)
- Source value
- 1.6
- Score
- 1.6
- Metric
- word error rate on VoxPopuli-Cleaned-AA
- Unit
- %
- Benchmark version
- AA-WER v2.0 per-dataset comparison, retrieved 2026-10-16
- Category
- audio
- Direction
- Lower is better
- Source type
- independent-evaluator
- Evaluator
- Artificial Analysis
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. Exact MAI-Transcribe-1.5 row on the cleaned VoxPopuli-AA dataset; this is one component, not the overall AA-WER result.
- Context
- The article does not report a per-row sample count. This dataset-specific result is not the overall weighted AA-WER score.
Artificial Analysis Earnings22-Cleaned-AA WER (Show measurement, test conditions, and source)
- Source value
- 4
- Score
- 4
- Metric
- word error rate on Earnings22-Cleaned-AA
- Unit
- %
- Benchmark version
- AA-WER v2.0 per-dataset comparison, retrieved 2026-10-16
- Category
- audio
- Direction
- Lower is better
- Source type
- independent-evaluator
- Evaluator
- Artificial Analysis
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Independent evaluation. Exact MAI-Transcribe-1.5 row on the cleaned Earnings22-AA dataset; this is one component, not the overall AA-WER result.
- Context
- The article does not report a per-row sample count. This dataset-specific result is not the overall weighted AA-WER score.
Microsoft-reported mean FLEURS WER across reported locales (Show measurement, test conditions, and source)
- Source value
- 4.86
- Score
- 4.86
- Metric
- mean word error rate across reported FLEURS locales
- Unit
- %
- Benchmark version
- Microsoft FLEURS chart, data as of July 2026; latest MAI-Transcribe-1.5
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- Microsoft AI
- Status
- active
- Retrieved at
- 2026-10-16
- Methodology
- Vendor-reported result. Microsoft reports mean WER across all reported FLEURS locales for the latest MAI-Transcribe-1.5 version; the page dates the data to July 2026. The source does not state the locale count or evaluation configuration in the visible chart description.
- Context
- Vendor-reported FLEURS value; it is a different dataset and setup from Artificial Analysis AA-WER, so it is not directly comparable with the 2.4% AA-WER result.
Microsoft FLEURS top-25-language WER (Show measurement, test conditions, and source)
- Source value
- 3.7
- Score
- 3.7
- Metric
- Word error rate (WER), mean across the top 25 global languages in the FLEURS benchmark
- Unit
- %
- Benchmark version
- Microsoft-reported FLEURS top-25-language aggregate; source states data updated as of 2026-08-12
- Category
- audio
- Direction
- Lower is better
- Source type
- vendor-reported
- Evaluator
- Microsoft Tech Community, Azure AI Foundry Blog
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Vendor-reported result. The primary article identifies the exact model as MAI-Transcribe-1.5 and reports FLEURS WER over the top 25 global languages. It does not publish a serving-build identifier, decoding settings, word/sentence denominator, or per-language cells.
- Context
- This is a vendor-reported top-25-language FLEURS aggregate, not an independent evaluation. Microsoft does not publish N, a confidence interval, the language-level cells, or decoding settings here. It is a different stated cohort from the existing 4.86% mean across all reported locales; do not treat the two aggregates as interchangeable or combine them.
PERIT-STT English WER, MAI-Transcribe-1.5 (Show measurement, test conditions, and source)
- Source value
- 1.57
- Score
- 1.57
- Metric
- pooled normalized word error rate (WER)
- Unit
- %
- Benchmark version
- PERIT-STT English real-world leaderboard, captured 2026-10-17
- Category
- audio
- Direction
- Lower is better
- Lower confidence bound
- 1.37
- Upper confidence bound
- 1.8
- Source type
- independent-evaluator
- Evaluator
- PERIT-STT, Perit
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact PERIT-STT entry microsoft/mai-transcribe-1.5. English real-world speech-to-text leaderboard, pooled normalized word error rate against two independent human references with senior adjudication; 22 systems share the held-out audio. The method page states three runs per clip, but does not publish the clip count or how repeated runs are aggregated.
- Context
- The model-specific number of clips is not published, so no sample count is recorded. The method describes three runs per clip; repeat aggregation is unspecified. The 95% interval is reproduced as published.
Nyra Verbatim English, Disfluency F1, MAI-Transcribe-1.5 (Show measurement, test conditions, and source)
- Source value
- 84
- Score
- 84
- Metric
- Disfluency F1
- Unit
- %
- Benchmark version
- Nyra Verbatim Speech Benchmark, README-Commit e2d1200537e8f66ccd3b0d800e9ba95c61c71d15
- Category
- audio
- Direction
- Higher is better
- Sample
- 4957
- Source type
- independent-evaluator
- Evaluator
- Nyra Health Verbatim Speech Benchmark
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact MAI-Transcribe-1.5 (Microsoft, mai-transcribe-1.5) table row, English. n=4,957 scored cached predictions. The table is sorted by Disfluency F1; Cutoff F1 is cutoff_lenient_f1.
- Context
- Nyra Health publishes this benchmark and also compares its own CrisperWhisper model. The Microsoft endpoint revision and a confidence interval are not published. The 0% Vocal Sound F1 in the same row is a reported score, not a missing measurement.
Nyra Verbatim English, Filler F1, MAI-Transcribe-1.5 (Show measurement, test conditions, and source)
- Source value
- 95.8
- Score
- 95.8
- Metric
- Filler F1
- Unit
- %
- Benchmark version
- Nyra Verbatim Speech Benchmark, README-Commit e2d1200537e8f66ccd3b0d800e9ba95c61c71d15
- Category
- audio
- Direction
- Higher is better
- Sample
- 4957
- Source type
- independent-evaluator
- Evaluator
- Nyra Health Verbatim Speech Benchmark
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact MAI-Transcribe-1.5 (Microsoft, mai-transcribe-1.5) table row, English. n=4,957 scored cached predictions. The table is sorted by Disfluency F1; Cutoff F1 is cutoff_lenient_f1.
- Context
- Nyra Health publishes this benchmark and also compares its own CrisperWhisper model. The Microsoft endpoint revision and a confidence interval are not published.
Nyra Verbatim English, Vocal Sound F1, MAI-Transcribe-1.5 (Show measurement, test conditions, and source)
- Source value
- 0
- Score
- 0
- Metric
- Vocal Sound F1
- Unit
- %
- Benchmark version
- Nyra Verbatim Speech Benchmark, README-Commit e2d1200537e8f66ccd3b0d800e9ba95c61c71d15
- Category
- audio
- Direction
- Higher is better
- Sample
- 4957
- Source type
- independent-evaluator
- Evaluator
- Nyra Health Verbatim Speech Benchmark
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact MAI-Transcribe-1.5 (Microsoft, mai-transcribe-1.5) table row, English. n=4,957 scored cached predictions. The table is sorted by Disfluency F1; Cutoff F1 is cutoff_lenient_f1.
- Context
- Nyra Health publishes this benchmark and also compares its own CrisperWhisper model. The Microsoft endpoint revision and a confidence interval are not published. 0% is the published table value.
Nyra Verbatim English, Cutoff F1, MAI-Transcribe-1.5 (Show measurement, test conditions, and source)
- Source value
- 83.3
- Score
- 83.3
- Metric
- lenient cutoff F1
- Unit
- %
- Benchmark version
- Nyra Verbatim Speech Benchmark, README-Commit e2d1200537e8f66ccd3b0d800e9ba95c61c71d15
- Category
- audio
- Direction
- Higher is better
- Sample
- 4957
- Source type
- independent-evaluator
- Evaluator
- Nyra Health Verbatim Speech Benchmark
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact MAI-Transcribe-1.5 (Microsoft, mai-transcribe-1.5) table row, English. n=4,957 scored cached predictions. The table is sorted by Disfluency F1; Cutoff F1 is cutoff_lenient_f1.
- Context
- Nyra Health publishes this benchmark and also compares its own CrisperWhisper model. The Microsoft endpoint revision and a confidence interval are not published. The cutoff score is the lenient cutoff_lenient_f1 variant, not the strict marker-based measure.
Nyra Verbatim English, Repetition F1, MAI-Transcribe-1.5 (Show measurement, test conditions, and source)
- Source value
- 86.8
- Score
- 86.8
- Metric
- Repetition F1
- Unit
- %
- Benchmark version
- Nyra Verbatim Speech Benchmark, README-Commit e2d1200537e8f66ccd3b0d800e9ba95c61c71d15
- Category
- audio
- Direction
- Higher is better
- Sample
- 4957
- Source type
- independent-evaluator
- Evaluator
- Nyra Health Verbatim Speech Benchmark
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. Exact MAI-Transcribe-1.5 (Microsoft, mai-transcribe-1.5) table row, English. n=4,957 scored cached predictions. The table is sorted by Disfluency F1; Cutoff F1 is cutoff_lenient_f1.
- Context
- Nyra Health publishes this benchmark and also compares its own CrisperWhisper model. The Microsoft endpoint revision and a confidence interval are not published.
Page 1 of 4
Popularity compared
How often the model is searched for, used and mentioned, compared with every model of its category.
Popularity index
0 to 100, higher is more popular
20 of 83 models
pyannote Community-1, pyannote: 73
Nemotron-3-Diarization, NVIDIA: 57
Gemini 3.5 Transcribe, Google: 52
Confucius4-R2T2, NetEase Youdao: 52
Granite Speech 5.0 470M TurboCTC, IBM: 52
Kokoro-82M, hexgrad: 51
Nova-3, Deepgram: 51
Fish Audio S2 Pro, Fish Audio: 51
NVIDIA Nemotron 3.5 Transcribe, NVIDIA: 50
Granite Speech 5.0 470M TurboCTC NC, IBM: 50
Orukeet, Oruk: 49
Lyria 3.5, Google: 47
Gemini 3.5 Live Translate, Google: 46
Phonon-2, Fermion Research: 44
Aura-2, Deepgram: 44
Gemini 3.1 Flash TTS, Google: 44
MiniMax-Music3, MiniMax: 41
Audio8-ASR-Infinite, Edge0 AI: 41
GPT-Realtime 2.1, OpenAI: 41
MAI-Transcribe-1.5, Microsoft: 0
MAI-Transcribe-1.5
As of November 2026 · Source: gradually.ai popularity index
Over time
Weekly values from October 18, 2026 to November 5, 2026
MAI-Transcribe-1.5Amazon Nova 2 SonicEleven Multilingual STS v2Eleven Music v2.5Eleven Text to Sound v2
All specifications
Architecture
Input and output
Watermarking and provenance
AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.
- AI watermarking
- Unknown
- no source
- Watermarking technique
- Not known
- no source
Licence and access
API
Further details
Prices
- Price
- $0.36 per hour of audioSource
Sources and data date
Show all 22 sources
Benchmark sources at a glance
Which independent evaluators tested this model. This is coverage, not a quality ranking.
| Evaluator | Test series | Composite indices | Recorded active measurements |
|---|---|---|---|
| Artificial Analysis | 0 | 0 | 1 |
Without vendor claims and historical or disputed results; those stay labelled in the single values. 1 measurements are not yet assigned to a test series.